Era of Working with Robots Draws Near: Physical AI 'Infrastructure Companies' Also Surge

More diverse corporate information on the startups mentioned in this article can be viewed at Unicorn Factory's big data platform 'Data Lab'.

As the global physical AI (artificial intelligence) industry grows rapidly, related infrastructure markets such as learning data construction, spatial recognition solutions, and energy and power management solutions are also expanding. In Korea, startups are increasingly entering the physical AI infrastructure market. These companies are targeting the market with B2B solutions that connect physical AI firms and traditional manufacturing industries.

Bound4 is a data foundry company focused on building data infrastructure to enable physical AI to be stably learned and operated in industrial settings. To address the limitations of simulation-based data learning, Bound4 offers a foundation data contract manufacturing service called "Foundry," which refines and provides real-world data.

It also established "FSD (Full Self-DataOps)," a fully autonomous data operation system that enables Physical AI to acquire and learn from its own data. In addition, key solutions include "Spatial AI," the execution engine for Physical AI, and "DropAI," a general-purpose data processing platform.

Bound4 is building training data for autonomous driving robots together with Naver Labs. Naver Labs has stated that it used Bound4's solution to reduce the time required for data creation from 90 minutes to 10 minutes and secured approximately 160,000 high-quality images over six months.

Hwang In-ho, CEO of Boundfor, stated, "The data for physical AI must be structured to learn the constraints of the real world directly, unlike the data used in language models," and added, "We will provide infrastructure that enables the stable accumulation and utilization of actual environmental data, making it possible to implement high-performance physical AI."

Boundfor provides data cleaning and processing solutions for physical AI learning, while Mobiletech possesses spatial infrastructure technology that enables physical AI to understand and utilize real-world spaces by implementing them in digital environments. Its core competitiveness lies in precisely digitizing physical spaces based on 'calibration' technology that corrects errors in sensor equipment, as well as 3D spatial information and digital twin technologies.

Mobile Tech provides three solutions — Sensor Suite, Calibration Suite, and Replica City — to enable autonomous mobile robots (AMRs) and self-driving vehicles to accurately perceive their surroundings and determine their location even in complex unstructured environments. Through these solutions, the company supports the construction of virtual spaces similar to real-world environments and facilitates the acquisition of diverse driving and operational data.

In particular, its strength lies in having established a data pipeline based on OpenUSD, the global industrial standard, securing high compatibility and scalability. Mobile Tech also supplies data to NVIDIA's digital twin platform, Omniverse, and its AI development platform, Cosmos. Domestically, it has secured 248 companies and institutional investors as clients, including Hyundai Motor and Naver.

Mobitech is expanding its business scope beyond physical AI into autonomous driving, national defense, and smart cities. Having already secured more than 100 patents and completed 923 projects to demonstrate its technological capabilities, the company plans to accelerate its global market penetration and technology advancement based on these achievements.

Meanwhile, Mobile Tech has raised a cumulative investment of 40.2 billion won, including a pre-IPO (pre-listing fundraising) round. The company is also moving forward with a special listing for technology companies based on its accumulated spatial information and digital twin technologies.

To ensure physical AI robots operate stably for long hours in logistics centers or manufacturing plants, minimizing downtime caused by charging is a key challenge. E-Tile Electronics is focusing on solving this problem through its wireless robot charging solution, "ELYX," which applies magnetic resonance-based wireless power transmission technology. The company's strategy is to build power infrastructure at industrial sites where robots are deployed, creating synergies with physical AI companies.

Eta Electronics' technology is designed to allow robots to receive power without any separate physical contact during operations. Unlike existing methods that require robots to stop at designated locations for charging, this system enables robots to be powered while moving through charging zones or along their operational routes. The company explains that the ability to perform work and charge simultaneously reduces downtime and improves operational efficiency.

Eta Electronics recently became the first domestic company to obtain KC certification in the field of wireless robot charging and has also passed certification by the U.S. Federal Communications Commission (FCC). Having demonstrated technical stability and reliability externally, the company plans to expand its wireless power infrastructure projects targeting the global physical AI market in the medium to long term.

[Money Today Startup Media Platform Unicorn Factory]